PackForcing：短视频训练足以支持长视频采样和长上下文推理
PackForcing: Short Video Training Suffices for Long Video Sampling and Long Context Inference
AI总结：
本文提出PackForcing框架，通过三部分KV缓存策略有效管理生成历史，实现高效内存使用和长视频生成，展示短视频监督足以支持高质量长视频合成。
摘要：
自回归视频扩散模型在长视频生成中表现出色，但受限于不可行的线性KV缓存增长、时间重复和累积误差。为解决这些问题，我们提出了PackForcing，一个统一框架，通过新颖的三部分KV缓存策略高效管理生成历史。具体而言，我们将历史上下文分为三种类型：(1) 沉底标记，保留早期锚帧以保持全局语义；(2) 中间标记，通过双分支网络融合渐进3D卷积与低分辨率VAE重编码，实现大规模时空压缩（32倍标记减少）；(3) 近期标记，保持全分辨率以确保局部时间一致性。为严格限制内存足迹而不牺牲质量，我们引入动态top-k上下文选择机制用于中间标记，并结合连续时间RoPE调整，无缝重新对齐因丢弃标记导致的位置间隙，且无显著开销。借助这种原理性的层次上下文压缩，PackForcing可在单个H200 GPU上生成连贯的2分钟832x480视频，以16 FPS运行。它实现了有限的KV缓存仅为4 GB，并能实现显著的24倍时间外推（5秒到120秒），在零样本或仅训练5秒片段的情况下均能有效运行。在VBench上的广泛结果展示了最先进的时间一致性（26.07）和动态度（56.25），证明短视频监督足以支持高质量长视频合成。
英文摘要：
Autoregressive video diffusion models have demonstrated remarkable progress, yet they remain bottlenecked by intractable linear KV-cache growth, temporal repetition, and compounding errors during long-video generation. To address these challenges, we present PackForcing, a unified framework that efficiently manages the generation history through a novel three-partition KV-cache strategy. Specifically, we categorize the historical context into three distinct types: (1) Sink tokens, which preserve early anchor frames at full resolution to maintain global semantics; (2) Mid tokens, which achieve a massive spatiotemporal compression (32x token reduction) via a dual-branch network fusing progressive 3D convolutions with low-resolution VAE re-encoding; and (3) Recent tokens, kept at full resolution to ensure local temporal coherence. To strictly bound the memory footprint without sacrificing quality, we introduce a dynamic top-$k$ context selection mechanism for the mid tokens, coupled with a continuous Temporal RoPE Adjustment that seamlessly re-aligns position gaps caused by dropped tokens with negligible overhead. Empowered by this principled hierarchical context compression, PackForcing can generate coherent 2-minute, 832x480 videos at 16 FPS on a single H200 GPU. It achieves a bounded KV cache of just 4 GB and enables a remarkable 24x temporal extrapolation (5s to 120s), operating effectively either zero-shot or trained on merely 5-second clips. Extensive results on VBench demonstrate state-of-the-art temporal consistency (26.07) and dynamic degree (56.25), proving that short-video supervision is sufficient for high-quality, long-video synthesis. https://github.com/ShandaAI/PackForcing
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
